Introduction

In this chapter, I shall turn from a consideration of anaphoric production to an analysis of anaphoric resolution in Chinese conversation. I shall demonstrate that anaphoric resolution in Chinese conversation can largely be accounted for by the interaction of the I- and M-principles. Unlike most previous studies of discourse anaphora, which are based on written narrative, the analysis of anaphoric production in the last chapter and that of anaphoric resolution in this chapter are based on naturally occurring conversation. Although conversation introduces some additional complexities, it also offers some extra means to confirm (or disconfirm) an analysis. Therefore, I shall end this chapter with a brief illustration of how the analysis made here conforms to what the participants in conversation are actually oriented to.

Anaphoric resolution in conversation

There appear to be some precise parallels between intrasentential and discourse anaphora in Chinese: for an antecedent of an anaphoric expression in Chinese conversation, a local subject is preferred to a local object, and a non-split antecedent, to a split one. If none of these NPs seems to be the possible candidate, the next, more remote utterance in the conversation is examined for possibilities in the same order, and so on until the most remote utterance in the conversation is reached. Failure to locate a conversation-internal antecedent leads to an inference to an exophoric or arbitrary interpretation (cf. sections 6.2 and 6.6 of chapter 6).

Let us now look at some illustrations of how anaphoric expressions in Chinese conversation are interpreted by the I- and M-principles. Consider first (8.1).

Most of the basic points concerned with roles and relations were made in the previous chapter. This chapter looks at them in more detail.

Agent and Patient

There are two issues concerning the roles Agent and Patient, first, the question whether they are universal, and, secondly, their relation to the meaning-based notional roles.

The universality of the distinction

It might seem to be obvious that all languages must make a grammatical distinction between different roles such as that of Agent and Patient, because, if there are two arguments with a predicator, it is essential to know which role is played by each of the two arguments. If, for instance, we are talking about someone hitting someone else, we need to know who does the hitting (the agent) and who was hit (the patient), and it is precisely that distinction that is communicated by the formal markers of Agent and Patient (1.2.2). Without such identification, it might be thought, communication would be impossible. Are there, then, languages that do not make the distinction at all, i.e. that do not grammaticalize basic notional roles such as agent and patient?

It is unwise to maintain that there cannot possibly be such languages, for it is dangerous to speculate about what must be in language, and all too often what appears to be an ‘obvious’ fact about language turns out to be merely a feature of English and familiar (usually European) languages.

As the title suggests, this book is a typological study of grammatical roles, such as Agent, Patient, Beneficiary, and of grammatical relations, such as Subject, (Direct) and Indirect Object, which are familiar concepts in traditional grammars; in addition it is concerned with the devices, such as the passive, that alter or switch (or ‘remap’ – see 1.1) the identities between such roles and relations. It will be apparent, however, in a typological study, that the grammatical systems of familiar languages are not typical of many of the languages of the world, and that the traditional terminology is inappropriate, as will be seen in the need to use such terms as ‘Ergative’, ‘Absolutive’, ‘Antipassive’ etc.

It should, nevertheless, be possible to suggest a consistent and reasonably simple overall framework within which such issues may be illustrated and discussed (though nothing is very simple in language). Yet very few attempts to do so have been made, and even fewer have been at all successful. The main aim of this book is to provide such a framework and to illustrate within it some of the typological characteristics of different languages. As such, it will not contain a great deal of theoretical discussion, though theoretical issues cannot be wholly ignored, for the framework must rest on certain theoretical assumptions and observations. One simple point, however, should be made: a typological study is concerned with similarities and differences between languages, and does not rest upon the assumption that there are universal (and identical) features across languages (see Palmer 1986: 2–3 and, for a more detailed theoretical discussion, Croft 1991: 17–32).

This chapter investigates in much more detail issues of syntax raised briefly in 1.3.1 and 3.1.

Syntactic pivots

Some of the grammatical relations, particularly the Subject in accusative languages, are involved in certain grammatical constructions. Thus, in English, if two sentences are coordinated, usually with and, the subject of the second is normally omitted or ‘deleted’, if it is coreferential with that of the first, as in (with the omitted NP shown in brackets):

1. The man came in. The man saw the woman.

2. The man came in and [the man] saw the woman.

This is only possible if both nouns are Subjects, as can be seen from the impossible sentences:

1. The man came in. The woman saw the man.

2. *The man came in and the woman saw [the man].

1. The man saw the woman. The woman came.

2. *The man saw the woman and [the woman] came in.

1. The man saw the woman. The boy heard the woman.

2. *The man saw the woman and the boy heard [the woman].

This restriction on deletion in coordination, together with similar syntactic conditions that are dealt with in more detail in later sections of this chapter, is treated in terms of ‘pivot’, the Subject being said to be the pivot for deletion in coordination in English. However, two NPs are involved in the coreferentiality and it may be better to refer to both of them as ‘pivots’, the first being the ‘controller’ and the second the ‘target’ (Foley and Van Valin 1985: 305), though in much of the discussion the term is used to refer to the second, the target.

The passive was briefly discussed in 1.4. In this chapter and the next various kinds of passives and similar constructions are examined in more detail.

The identification of the passive

1. A simple and obvious type of passive is illustrated by (see 1.4.1):

2. The policemen caught the thief

3. The thief was caught by the policemen

Commonly the Subject of the active sentence is omitted in the passive:

1. The boy was hit

Sentences such as these are referred to as ‘agentless passives’. It is important to note that in English it is not possible simply to omit the Subject of the active sentence to produce:

1. *Caught the thief

If the Agent is not to be mentioned the passive must be used.

If the passive is a typologically valid category, it must, like all such categories, be identified (i) in terms of the meaning or function it shares across languages, and (ii) in terms of its formal marking in individual languages. (For a discussion see Palmer 1986: 2–7.) It can be said that the basic functions of the passive are the promotion of the Patient (or non-Agent) and the demotion or deletion of the Agent (but see 6.7). This does not involve the acceptance of a particular syntactic theory, but merely implies that the functions of the roles of Agent, Patient etc. in the passive can be accounted for in terms of variation from their functions in the active sentence.